Post-increment floating-point loads and stores

[andreaskurth]

Is it correct that the post-increment extension currently only supports integer loads and stores but not floating-point loads and stores? If so, are there any plans for enabling that extension for floating-point loads and stores? That would help increase the utilization of the FPU (depending on the hot loop even very significantly) and therefore floating-point performance.

[Silabs-ArjanB]

Hi @accuminium, your understanding is correct. There are no plans for CV32E40P to add a similar extension for floating-point loads and stores. (I would assume though (but did not try it out) that you get what you want when you would use the PULP_ZFINX = 1 parameter together with integer load/stores).